As an approach to and method for the analysis of set relations, Qualitative Comparative Analysis (QCA) belongs in the toolbox of many social scientists (and organizational and management researchers alike). Although it has proven to be very popular, a question that one still often confronts is: what is QCA? Even those who understand the technicalities of QCA wonder what the qualitative part in QCA represents in practice. When looking for answers to both questions, we find a variety of responses, some of which have more or less subtle, but important implications.
A definition from the founder
When talking about QCA, the first place to look for an answer is, of course, the work of the founder of QCA, Charles Ragin. His first book, The Comparative Method, offered the definition of QCA as the “comparison of wholes as configurations of parts” (p. 84). In contemporary QCA language, “parts” are the constitutive conditions of configurations; “wholes” I understand to be the cases under analysis. The cases in turn are members of conjunctions, which were also referred to as “types” in Ragin’s work. Following this short definition alone, this means we are doing QCA as long as we are comparing cases described by conjunctions.
Sounds good, but…
However, this definition does not seem to apply to what is usually practiced under the label of QCA, where two issues stand out. First, the search for necessary conditions does not seem to fit under this definition. Under the usual procedure, we are looking for a single necessary condition or the union of SUIN conditions which do not qualify as a configuration. However, recent developments take a more comprehensive approach toward necessary-condition analyses and also allow for the intersection of conditions, i.e., conjunctions. (See Bol/Luppi in the PRQ QCA symposium and Thiem/Dusa’s QCA package.) In this view, inquiries into necessity also fit the definition of QCA. (How meaningful this is is another matter. In my reading, adding conjunctions to the picture only serves to boost the relevance (i.e., coverage) of the necessary term and is primarily a parameter-driven approach.)
The second problem is that QCA is not just comparing cases and configurations for discerning sufficient terms of an outcome. If this were all, it would not be distinguishable from typological theory as another set-relational approach (see Bennett/Elman on QCA and typological theory). What sets QCA and typological theory apart is that the former invokes algorithms for the identification and elimination of redundant conditions (more on algorithms in a minute). For this reason, any definition of QCA should take this into regard. Taken together, finding a definition for QCA leaves one in a somewhat uncomfortable position because inquiries into necessary terms do not require an algorithm identifying redundant conditions (at least, that’s not how necessary-condition analyses are implemented at present).
One QCA, many algorithms
In recent years, Michael Baumgartner put into focus that QCA as invented by Ragin relies on the Quine-McCluskey (QMc) algorithm. He proposes Coincidence Analysis (CNA) as an alternative to QMc because CNA solves some problems that QMc supposedly has (the details of the algorithm controversy are not relevant here). What is important is that Baumgartner often criticizes QCA when he actually means the QMc algorithm. The understanding of QCA as the comparison of wholes as configurations of parts is pivotal here. QMc and CNA rely on different algorithms, but they have in common that they compare configurations with the goal of identifying redundant conditions. My preferred view on the QMc vs CNA debate is that QCA subsumes a class of algorithms and is not linked to one specific of them. An equivalent in quantitative research would be regression that includes many different estimation techniques such as OLS and GLS. Questioning one particular estimation technique does not entail questioning regression altogether.
What QCA is absolutely not
In its early years, the implementation of QCA was (and sometimes still is) justified by the availability of a limited number of cases that do not allow it to run a statistical analysis (usually meant to be some form of frequentist analysis). This choice was based on the understanding of QCA as a medium-n method located in between case studies as a small-n method and regression as a large-n method. However, this has never been and will never be a good reason for doing QCA because it makes the choice between a regression analysis and a set-relational method conditional on the number of cases. Since the inferences that one derives from regression analysis and QCA are fundamentally different, one should beware of putting the cart before the horse. One should first decide whether one is committed to the underlying assumptions of regression or QCA and then live with the number of cases that are available for analysis.
And what about the “Q” in QCA?
It is easy to answer the question about the “Q” in QCA by drawing on the extremely useful distinction between QCA as an approach and QCA as a method (see chapter 1 in edited volume by Ragin/Rihoux). QCA as an approach draws on qualitative case knowledge at various points of the analysis, which is the spirit in which it has been originally proposed by Ragin. QCA understood as a method does not require case knowledge, but boils down to the processing of data (i.e., the truth table) with an algorithm. In this instance, the “Q” indeed loses its meaning and something like Set-Relational Comparative Analysis would be the better term.
So what now?
Even a short digression into the question of what QCA is shows that it is more difficult to answer than one might have imagined. One might discard all of this as definitional subtleties that do not matter in the end. This is probably true from an applied side, but the algorithm debate for example points to the importance of having an understanding about the nature of QCA. QCA is constantly criticized for a variety of reasons (in particular in these days) and one should not give the impression of an adding another point to the list when the discussion is not about QCA and set-relational analysis per se, but “only” about the proper algorithm.
A working definition
Of course, I cannot end this post without offering a definition of QCA. Given what I have written, a definition that works is “QCA is a case-based approach that searches for minimally sufficient terms and maximally relevant necessary terms”. This definition basically works backwards from the term and common practice in empirical research. The attribute “case-based” is needed because of the “Q”; “minimally sufficient” implicitly introduces an algorithm to the definition; and “maximally relevant” refers to conjunctions (“wholes”) in analyses of necessity. It might not be an overly elegant definition, but it subsumes all elements that are needed in light of what is commonly understood as QCA. The alternative would be to come up with another label for what is referred to as QCA which, however, is unlikely (and would be unfortunate) because QCA established itself as a brand name in the field.